Training Sessions 3 and 4: Introduction to SEER, Medicare and SEER-Medicare

Transcription

1 Training Sessions 3 and 4: Introduction to SEER, Medicare and SEER-Medicare Rodney Sparapani, PhD The Center for Patient Care and Outcomes Research and the Division of Biostatistics Medical College of Wisconsin April 20 and May 4, 2012

2 Training Outline 1. What is SEER? 2. What is Medicare? 3. What is SEER-Medicare? 4. Comorbidity via Medicare Claims 5. Breast Cancer Identification via Medicare Claims 6. Conclusions If all else fails, read the instructions. - Donald Knuth, renowned computer scientist

3 Surveillance Epidemiology and End Results (SEER) Free download Docs Ries, Pollack and Young. Cancer patient survival: Surveillance, Epidemiology, and End Results program, Journal of the National Cancer Institute 1983:70; all incident cancers in certain registry areas (except non-melanoma skin cancer): de-identified, i.e. no names, SSN, etc. demographics, Census 2000, mortality/cause of death, and up to 10 lifetime primary cancer episodes cancer info: diagnostic/prognostic, initial treatment but not chemotherapy nor endocrine therapy

4 Surveillance Epidemiology and End Results (SEER) (cont.)

5 Medicare US national health insurance program for the disabled and elderly (65+) managed by the Centers for Medicare and Medicaid Services (CMS) Standard Analytic Files (SAF) created for research Part A: claims for hospital/inpatient stays (MEDPAR/Inpatient SAF: ICD-9-CM diagnoses and procedures) Part B: claims for physician/clinic visits (Carrier and Outpatient SAFs: ICD-9-CM diagnoses and HCPCS procedures) Part C: claims for Medicare Advantage AKA HMO (A/B claims NA but D are available) Part D: claims for prescription drugs (PDE SAF: NDC, generic and brand names)

6 Medicare and the Elderly Aged 65+ (circa ) Part A: 98% enrolled; in general, no monthly premium Part B: 94% enrolled; monthly premium Part C: 20% enrolled; covers Part A/B/D services in a managed care setting Part D: 58% enrolled (includes Part C: PDE claims are available) 17% enrolled in Retiree Drug Subsidy (claims NA) 14% not enrolled, but have Creditable Coverage (claims NA: Part D equivalent coverage) Total: 89% have drug coverage Neuman and Cubanski. Medicare Part D Update: Lessons Learned and Unfinished Business. NEJM 2009:361;

7 Medicare Demographics and Eligibility/Enrollment Beneficiary Summary File: CCW.DENOM A/B/C/D eligibility/enrollment information Birth date and death date, if any Age: age=%ageyr(birth_dt, 01JAN2008 d); Gender Race (B/W/O/API/H/NA) format race race.; Research Triangle Institute (RTI) Race Imputation RTI Race medicareindicators/medicareindicators2.htm Address, City, County, State and ZIP code

8 Medicare Part A Hospital Stays, 2 flavors: MEDPAR and Inpatient SAF (CCW.IP) Also Skilled Nursing Facility (SNF), Home Health Agency (HHA) and Hospice Inpatient SAF has more info than MEDPAR: mainly physician ID Both have Hospital ID, dates, charges, etc. Hospital Stays have up to 10 ICD-9-CM diagnosis and 6 ICD-9-CM procedure codes

9 Medicare Part B 2 flavors: Carrier SAF and Outpatient SAF: CCW.CO Carrier SAF is for physician billing: CCW.CARRIER Outpatient SAF is for clinic billing: CCW.OP up to 10 ICD-9-CM diagnosis codes and 1 HCPCS procedure code Both have physician ID (unfortunately, only Outpatient has hospital ID while Carrier are more numerous) UPIN used up to June 2007; replaced by NPI nppes.cms.hhs.gov/nppes/npiregistryhome.do neither are unique; only the AMA ResearchID is unique with approval for research, you can buy from the AMA with a ResearchID to UPIN/NPI crosswalk

10 Medicare Part C Medicare Advantage AKA Health Maintenance Organization (HMO) Part A and B claims appear in their respective places But, HMOs do not provide all of their claims to CMS since there is no additional compensation Therefore, we often omit Part C patients from our cohorts since their treamtents and services are incomplete However, all Part D claims are provided to CMS

11 Medicare Part D Part D Event SAF: CCW.PDE Enrollment began January 2006, but not compulsory until June 2006 National Drug Code (NDC), generic/compound name, brand name fill date, pill count, charges, etc. Part C enrollee claims available Part D Retiree Drug Subsidy claims not available Non-part D Creditable Coverage claims not available

12 Medicare and HIPAA Health Insurance Portability and Accountability Act (HIPAA) patients are identified, i.e. names, SSN, etc. cannot easily be de-identified, i.e. Medicare includes the dates served by, and the identifiers of, hospitals, clinics, physicians, etc. Research practices and security dictated by Data Use Agreement (DUA) HIPAA not applicable to Medicare claims for research purposes, but DUA has similar requirements PCOR SOP 002: Good Protected Health Information Practices

13 Medicare Resources Medicare Chronic Condition Warehouse (CCW) Documentation Research Data Assistance Center (ResDAC) Medicare Documentation SES Part D Study: /ses/doc/medicare Diagnosis/Procedure codes (except CPT): /ses/doc/medicare/code Reference Sets_2008.xls

14 Medicare Coding Resources International Classification of Disease, ICD-9-CM: ICD-10-CM: American Medical Association (AMA) Current Procedural Terminology (CPT) AKA HCPCS Level I: code books in MCW library and PCOR Health Care Procedure Coding System HCPCS Level II: 02_HCPCS_Quarterly_Update.asp National Drug Code (NDC): ndc/default.cfm

15 SEER-Medicare http: //healthservices.cancer.gov/seermedicare combines SEER ( ) with Medicare claims ( ; Part D ): adds treatment information like chemotherapy and endocrine therapy we focus on SEER ( ) with Medicare claims ( ) includes a 5% Medicare non-cancer control group de-identified: no names, SSN, address, etc. no other identifiers like hospital, clinic, physician, etc. dates of service are available Patient Entitlement and Diagnosis Summary File (PEDSF) /seermedicare/doc/new/pedsf.pdf

16 SEER-Medicare and Breast Cancer (ICD-10: C50) We focus on the patient s first lifetime cancer, i.e. dx_dt1 and nearly all other variables that end in 1 (created variables brstins, brstmal and erpr do not) Patients who s first lifetime cancer is a breast cancer are identified as follows: in situ (Stage 0) breast cancer brstins=1 invasive (Stage 1-4) breast cancer brstmal=1 Diagnostic/prognostic variables: Extension of Disease (EOD): TNM Stage Classification Tumor Grade grade1 Tumor Markers: Estrogen Receptor (ER) and Progesterone Receptor (PR) ER and PR combined into one variable erpr

17 SEER-Medicare and Extension of Disease (EOD) American Joint Committee on Cancer (AJCC) staging: 0 (in situ), 1-4 (invasive) and 4 is metastatic TNM Stage Classification: T for primary Tumor, N for regional lymph Nodes and M for distant Metastasis 4 variables: T, tumor size (e10sz1); N, nodal involvement (e10nd1) and positive nodes (e10pn1); and M, tumor extension (e10ex1) Generally, the TNM components are 3 separate variables (T, N and M) rather than Stage (0-4) due to missingness (TX, NX and MX) i.e. if the regional lymph nodes were not removed, then N is unknown (NX) For breast cancer, page 120 of http: //seer.cancer.gov/manuals/eod10dig.pub.pdf /seermedicare/doc/new/eod10dig.pub.pdf

18 SEER and Missing Values SEER variables are somewhat consistently coded for missingness For example, ER is 0 for Not Done, 3 for Borderline, 8 for Not in Chart and 9 for Unknown For our purposes, these have all been recoded to a SAS missing value._ See /seermedicare/sas/new/pedsf.sas which creates BREAST.PEDSF

19 SEER and Missing Values with Informats *example based on /seermedicare/sas/new/pedsf.sas; proc format; invalue _0389_ 0, 3, 8, 9=._ 2 =0; *1 does not need re-coding; run; data erpr; input... tumor1_1 _0389_1. tumor2_1 _0389_1....;... if brstmal=1 brstins=1 then do; if tumor1_1 tumor2_1 then erpr=1; else if tumor1_1=0 & tumor2_1=0 then erpr=0; end; run;

20 Comorbidity Comorbid conditions: conditions unrelated to the disease being studied We follow the Charlson approach: Charlson et al. J Chron Dis 1987:40; With enhanced ICD diagnoses (see /ses/sas/ccw/enhanced.sas): Quan et al. Medical Care 2005:43; NCI Combined: combining inpatient and outpatient claims for the 4 major cancers; breast, colorectal, lung and prostate Klabunde et al. Ann Epidemiol 2007;17:584-90

21 Comorbidity Warehouses and SAS Programs We define comorbidities in time windows e.g. during the period starting one year prior to surgery up to the day before surgery OR from two years prior until surgery To facilitate these varying definitions: we construct static warehouses which need to re-built only when new Medicare data arrives (rarely) Inpatient: inmorbid.sas creates CCW.INMORBID Outpatient: outmorbid.sas creates CCW.OUTMORBID

22 Comorbidity Warehouses and SAS Programs CCW.INMORBID and CCW.OUTMORBID Non-cancer Diagnosis/Disease: 1. Acute MI, 2. Congestive Heart, 3. Peripheral Vascular, 4. Cerebrovascular, 5. Dementia, 6. Chronic Pulmonary, 7. Rheumatologic, 8. Peptic Ulcer, 9. Mild Liver, 10. Mild/moderate Diabetes, 11. Severe Diabetes, 12. Hemiplegia/Paraplegia, 13. Renal, 14. Moderate/Severe Liver, 15. AIDS, 16. Old MI

23 Comorbidity Warehouses and SAS Programs CCW.INMORBID and CCW.OUTMORBID Primary and Secondary Neoplasms System: 20. Oral/Pharynx, 21. Digestive, 22. Respiratory, 23. Bones/Joints, 24. Soft Tissue, 25. Melanoma, 26. Breast, 27. Genitals, 28. Leukemia, 29. Urinary, 30. Eye/Orbit, 31. Brain/Nervous, 32. Endocrine, 33. Lymphoma/Myeloma, 34. Misc.

24 Comorbidity Warehouses and SAS Programs: Bit Arithmetic Combined: comorbid.sas creates CCW.COMORBID from CCW.INMORBID and CCW.OUTMORBID Example 1: No Primary Breast Cancer diagnoses in June 2006 P =0 Example 2: Primary Breast Cancer diagnoses in June 2006 on the 1st and 15th P =32770= band(2**15, P )=32768= 2 15 band(2**10, P )=0

25 Medicare, Breast Cancer and Comorbidity Warehouses Combined: charlson.sas summarizes CCW.INMORBID and CCW.OUTMORBID: CCW.CHARLSON12 requires 12 months of prior claims and CCW.CHARLSON24 requires 12 months of prior claims and 12 months of confirmation claims 16 Non-cancer Diagnoses: C C Primary Cancers: P P Secondary Cancers: S S Also: /ses/sas/ccw/alg.sas summarizes CCW.INMORBID and CCW.OUTMORBID for identifying breast cancer patients (but that also needs surgeries) Once you have identified breast cancer patients: you can calculate their non-cancer comorbidities via NCI Combined

26 Medicare, Breast Cancer and Comorbidity Warehouses merge ccw.charlson12(keep=bene_id c c )...; by bene_id; %_array(array=_c, dim1=2001:2008, dim2=12, dim3=16, length=3, var=c); array _cwgt(16) _temporary_ ( ); if nmiss(_c(dx_yr1, dx_mo1, 1)) then nci=-1;

27 Medicare, Breast Cancer and Comorbidity Warehouses else do; if esrd then _c(dx_yr1, dx_mo1, 13)=1; if _c(dx_yr1, dx_mo1, 11) then _c(dx_yr1, dx_mo1, 10)=1; if _c(dx_yr1, dx_mo1, 14) then _c(dx_yr1, dx_mo1, 9)=1; nci=0; do i=1 to 16; if _c(dx_yr1, dx_mo1, i)>0 then nci=(nci+_cwgt(i)); end; end;

28 Identifying Breast Cancer with Medicare First Year Sens. Spec. PPV Medicare Author Claims McClish % Hospital only Cooper % Hospital only 93.6% Hospital/Physician Warren % 99.9% 91.3% Hospital only 76.2% 99.3% 36.3% Hospital/Physician Freeman % 99.86% 70.0% Hospital, Physician & Clinic Nattinger % 99.95% 89.0% Hospital, Physician & Clinic

29 Identifying Breast Cancer with Medicare Study PPV 95% CI Prior Cohort Year(s) Nattinger et al. 89.0% (86.7%, 91.6%) 3 Validation 88.1% (85.6%, 90.7%) 3 Training Survey 82.5% 1 CA 86.9% 4 FL/IL/NY 85.7% All

30 The Breast Cancer Identification Algorithm for Medicare Claims S1. negative if either no breast cancer diagnosis OR no breast cancer procedure; otherwise S2 S2. Both needed to be positive, go to S4; otherwise S3 mastectomy OR {a lumpectomy/partial mastectomy AND a radiotherapy claim with a breast cancer diagnosis} 2+ claims 31+ days apart with a primary diagnosis of breast cancer S3. All 3 needed to be positive, go to S4; otherwise negative a mastectomy OR a lumpectomy/partial mastectomy 2+ claims 31+ days apart with a diagnosis of primary or in situ breast cancer neither a secondary diagnosis of breast cancer nor a primary diagnosis of another cancer S4. negative if in prior 3 years: a history of breast cancer diagnosis OR {a breast cancer diagnosis AND a lumpectomy/partial mastectomy/mastectomy}

31 The Breast Cancer Identification Algorithm for Medicare Claims (cont.) Nattinger, Laud, Bajorunaite, Sparapani, Freeman. An algorithm for the use of Medicare claims data to identify women with incident breast cancer. HSR 2004:39; Here Tables 6 and 7 replace the original Table 1

32 The Breast Cancer Identification Algorithm for Medicare Claims (cont.) Table 6. Diagnosis Codes (SIMPLIFIED) Diagnosis ICD-9-CM Diagnosis Code(s) (Algorithm Step(s) Involved) Primary breast cancer (1, 2, 4) 174.x Breast carcinoma in situ (1, 2) Secondary breast cancer (3) 198.2, Non-breast cancer (3) 140.x , 230.x (except 174.x, 196.x, 198.2, , 232.5, 233.0, 238.3, 239.3) Breast cancer history (4) V10.3

33 The Breast Cancer Identification Algorithm for Medicare Claims

34 The Breast Cancer Identification Algorithm for Medicare Claims (cont.) Table 7. Procedure Codes (CORRECTED AND EXPANDED) Procedures ICD-9-CM Procedure CPT/HCPCS Code(s) (Algorithm Step(s), if any) Code(s) Tumor Biopsy (1) 85.1x 19000, 19001, , 19110, Lymph node biopsy/dissection only (1) (axillary), 40.3 (regional), (unspecified), (axillary) (axillary), (axillary), (axillary), (axillary) Lumpectomy (1-4) 85.20, , 19125, Partial mastectomy (1-4) 85.22, , 19162, 19301, Mastectomy (1-4) , , 19182, 19200, 19220, 19240, Radiotherapy (2) , , , , 77427, 77431, 77520, 77522, 77523, 77525, 77750, 77761, 77763, , Chemotherapy J8510, J8520, J8521, J8530-J9999 (except J9202, J9209, J9212-J9214, J9217, J9218) Lymph node dissection with other procedure Sentinel lymph node biopsy (axillary) (axillary), (axillary), (axillary & mammary), (axillary), (axillary), (axillary), (axillary & mammary), (axillary) A9520 (radioactive dye, ), A9541 (radioactive dye, 2006+), (axillary), (axillary), (injection), (injection)

35 Outcomes of older women with early stage breast cancer Updated Epi/PCOR presentation written by Rodney; generally, you can trust it, even when it conflicts with a publication: /survey/doc/survey.pdf Nattinger A, Pezzin L, Sparapani R, Neuner J, King T, Laud P. Heightened attention to medical privacy: challenges for unbiased sample recruitment and a possible solution. American Journal of Epidemiology 2010;172(6): PMC Neuner J, Yen T, Sparapani R, Laud P, Nattinger A. Fracture risk and adjuvant hormonal therapy among a population-based cohort of older female breast cancer patients. Osteoporosis International 2011;22: PMC

36 Outcomes of older women with early stage breast cancer (cont.) Kong A, Yen T, Pezzin L, Miao H, Sparapani R, Laud P, Nattinger A. Socioeconomic and racial differences in treatment for breast cancer at a low-volume hospital. Annals of Surgical Oncology 2011;18: PMC Yen T, Czypinski L, Sparapani R, Guo C, Laud P, Pezzin L, Nattinger A. Socioeconomic factors associated with adjuvant hormone therapy use in older breast cancer survivors. Cancer 2011;117(2): PMC Yen T, Sparapani R, Guo C, Neuner J, Laud P, Nattinger A. Elderly breast cancer survivors accurately self-report key treatment information. Journal of the American Geriatrics Society 2010;58(2): PMC

37 Outcomes of older women with early stage breast cancer (cont.) Yen T, Fan X, Sparapani R, Laud P, Walker A, Nattinger A. A contemporary, population-based study of lymphedema risk factors in older women with breast cancer. Annals of Surgical Oncology 2009;16(4): PMC